Extended Data Fig. 4: Interaction terms of statistical models used to test for buffering of pesticide hazard by semi-natural habitat (SNH) proportion in landscapes and for increased effects of pesticide hazard in bee-attractive crops.
From: Pesticides and habitat loss additively reduce wild bees in crop fields
a–d, Estimates ± 95% CI derived from linear mixed effects models accounting for non-independence within dataset (N = 681 sites) on the studied descriptors of bee communities in crop fields (MPD = mean pairwise distance) by pesticide hazard quotient (HQ, calculated from pesticide application protocols considering application rates and the toxicity of active ingredients to bees) × SNH (a), high pesticide use intensity intensity (based on production system considering typical application protocols) × SNH (b), HQ × bee attractiveness of focal crop (c), high pesticide use intensity × bee attractiveness of focal crop (d).
